TurboDeploy

Platform engineering is overhyped. For a team of 5, your entire "internal developer platform" should be: A Dockerfile. A GitHub Actions workflow. A README. That's it. That's the platform. I keep seeing startups with 3 engineers building Backstage portals, service catalogs, and "golden paths" before they even have 10 customers. You don't need a "developer experience layer." You need to ship. The irony is that most of these platform engineering efforts are designed to reduce complexity. But the platform itself becomes the complexity. I've seen teams spend 3 months building an internal deployment dashboard. Meanwhile, a GitHub Actions workflow that does the same thing takes 30 minutes to write and costs $0/month. Platform engineering makes total sense at 50+ engineers. When you have 10 teams deploying independently. When the cognitive load of understanding the infrastructure is genuinely slowing people down. At that scale, yes. Build the platform. But at 5 engineers? Everyone already knows the project. Everyone already knows the infrastructure. The "cognitive load" is minimal because there are only 3 services and one database. At that point, platform engineering isn't reducing complexity. It's creating it. The best platform is the one nobody maintains because it's boring enough to just work. A Dockerfile. GitHub Actions. A README. Push to main. Deployed in 4 minutes. Hot take? Or just common sense?

The 5 AWS services every startup actually needs. And the 5 you should skip entirely. After deploying dozens of apps to AWS, I've noticed a pattern. Startups consistently waste money on services they don't need while ignoring the ones that actually matter. Here's the "use" list: 1. ECS Fargate. ~$18/month. Serverless containers, zero ops. You write a Dockerfile, push it and it runs. No servers to manage. No patches. 2. RDS Postgres. $15/month for db.t4g.micro. Managed database with automated backups, failover and point-in-time recovery. Set it up once, forget about it. 3. S3. $1/month for most startups. File storage that scales infinitely. Every startup needs it. 4. CloudWatch. Free tier covers most early-stage needs. Logs and basic metrics without a third-party tool. 5. SES. 62,000 free emails per month. No need for SendGrid or Mailgun until you outgrow this. Total: roughly $35/month. Now here's the "skip" list: 1. EKS (Kubernetes). $75/month just for the control plane. Before you run a single container. Overkill for teams under 20 engineers. 2. DynamoDB. Sounds cheap until you hit provisioned capacity. Postgres does everything a startup needs. 3. Lambda for everything. Great for event-driven tasks. Terrible as a replacement for your entire backend. 4. Amplify. Vendor lock-in disguised as simplicity. Hard to migrate away from later. 5. Elastic Beanstalk. It was great in 2015. ECS has made it obsolete. The "use" stack costs $35/month. The "skip" stack costs $400+/month. And both run the exact same app. Full analysis with cost breakdowns in the comments. Which AWS services do you swear by? And which ones did you regret adopting?